JP6849723B2 - Methods and devices for generating information - Google Patents

Description

Examples of the present invention relate to the field of computer technology, specifically to methods and devices for generating information.

Currently, it is usually possible to mine entities in text using Named Entity Recognition (NER) techniques and Entity Linking (EL) techniques. In addition, NER can recognize proper nouns such as a person and a company. The EL can solve the problem of the same instruction of an entity by linking the word in the text and the entity in the knowledge picture book. However, it is currently not possible to recognize and link events.

The embodiments of the present invention provide methods and devices for generating information.

In a first aspect, the embodiments of the present invention provide a method for generating information. The method includes a step of receiving an object and a target text containing descriptive information about the object, a step of performing dependency parsing on the target text to generate a dependency tree of the target text, and at least one preset. A step of matching one parse structure tree with the dependency tree to obtain at least one triplet (triplet) consisting of a subject, a predicate, and an object, a word contained in the triplet of the at least one triplet, and the above. It comprises the step of determining one target triplet from the at least one triplet based on the preset weights of the parsed structure tree that are matched to obtain the triplet.

In some embodiments, from the at least one triplet, based on the words contained in the triplet of the at least one triplet and the preset weights of the syntactic structure tree that are matched to obtain the triplet. The step of determining one target triplet is to determine the quantifier and the modifier of the association in the target text based on the dependency tree, and the object to be modified by the quantifier and the object to be modified by the modifier of the association. A step of updating the at least one triplet based on the determined quantifier, the modifier and the object, and a step of determining one target triplet from the updated at least one triplet.

In some embodiments, the step of updating the at least one triplet based on the determined quantifier, adjunct modifier and object was determined for the triplet of the at least one triplet. Modifies the determined object in response to the step of determining whether the object matches the subject or object of the triplet and the determination that the determined object matches the subject of the triplet. The step of combining the quantity words, the modifiers to be used, and the subject of the triplet, and determining the combined text as the subject of the triplet, and the determination that the determined object matches the subject of the triplet. In response, it includes a step of combining a quantifier, a coalition modifier, and the subject of the triplet that modify the determined object, and determining the combined text as the subject of the triplet.

In some embodiments, from the at least one triplet, based on the words contained in the triplet of the at least one triplet and the preset weights of the syntactic structure tree that are matched to obtain the triplet. The step of determining one target triplet determines, for the triplet of at least one triplet, the preset weights of the syntactic structure tree that are matched to obtain the triplet and is included in the triplet. At least one of the steps of determining the number of characters of a word, determining the degree of co-occurrence of words contained in the triplet, and determining the score of the triplet based on the determined weight, the number of characters and the degree of co-occurrence. This includes the step of determining the triplet with the highest score as the target triplet.

In some embodiments, the method obtains a historical target obtained by matching a predetermined syntax structure tree of the at least one historical target triplet with the step of obtaining at least one historical target triplet. It further includes a step of statistics on the number of triplets and a step of determining the weight of the at least one syntactic structure tree based on the statistical result.

In some embodiments, the method comprises determining at least one historical event information associated with a target text in a preset historical event information set based on the target triplet, and the target text and at least said. It further includes a step of determining the similarity of one historical event information and a step of outputting the historical event information having the highest similarity with the target text.

In some embodiments, the historical event information includes participant information and trigger word information. In the step of acquiring at least one historical event information related to the target text in the historical event information set preset based on the target triplet, the subject or object of the target triplet is in the historical event information set. A step of determining whether or not a condition that matches the participant information of the history event information and a condition that the predicate of the target triplet matches the trigger word information of the history event information in the history event information set are satisfied. Includes a step of determining that the historical event information meets at least one of the above-mentioned conditions associated with the target text.

In some embodiments, the historical event information includes keywords. The step of determining the similarity between the target text and the at least one historical event information includes a step of acquiring a first word set by dividing the target text and a history of the at least one historical event information. A step of acquiring a second word set by concatenating each keyword included in the history event information with respect to the event information and dividing the concatenated text, and the first word set and the second word set. Includes a step of determining the similarity between the target text and the historical event information based on the word set of.

In a second aspect, the embodiments of the present invention provide an apparatus for generating information. The device is configured to receive an object and a target text containing descriptive information about the object, and to perform dependency parsing on the target text to generate a dependency tree for the target text. The dependency tree generation unit configured in is matched with at least one preset syntactic structure tree and the dependency tree to obtain at least one triplet consisting of a subject, a predicate, and an object. One from the at least one triplet based on the triplet determination unit and the preset weights of the syntactic tree matched to obtain the word contained in one triplet in the at least one triplet and the triplet. Includes a target triplet determination unit configured to determine the target triplet.

In some embodiments, the target triplet determination unit comprises a quantifier modifier configuration module configured to determine a quantifier and a quantifier in the target text based on the dependency tree, and the quantifier. An object determination module configured to determine the object to be modified and the object to be modified by the association modifier, and to update the at least one triplet based on the determined quantifier, association modifier and object. It includes a triplet update module that is modified and a target triplet determination module that is configured to determine one target triplet from at least one updated triplet.

In some embodiments, the triplet update module further determines, for the triplet of at least one triplet, whether the determined object matches the subject or object of the triplet. , In response to the determination that the determined object matches the subject of the triplet, combine the quantifiers, modifiers and the subject of the triplet that modify the determined object, and combine the combined text. A quantifier, a coalition modifier, and an object of the triplet that are determined as the subject of the triplet and that modify the determined object in response to the determination that the determined object matches the object of the triplet. Is combined, and the combined text is determined as the object of the triplet.

In some embodiments, the target triplet determination unit further determines, for the triplet of at least one triplet, the preset weights of the syntactic structure tree that are matched to obtain the triplet. Then, the number of characters of the word contained in the triplet is determined, the degree of co-occurrence of the word contained in the triplet is determined, the score of the triplet is determined based on the determined weight, the number of characters and the degree of co-occurrence, and the score of the triplet is determined. The triplet with the highest score among at least one triplet is configured to be determined as the target triplet.

In some embodiments, the device matches a given syntax structure tree of the at least one historical target triplet with a historical target triplet module configured to acquire at least one historical target triplet. A triplet quantity statistics module configured to statistic the number of historical target triplets obtained by letting, and a weighting module configured to determine the weight of at least one syntax structure tree based on the statistical results. Further includes a weight installation unit consisting of

In some embodiments, the device is configured to determine at least one historical event information associated with a target text in a preset historical event information set based on the target triplet. The determination unit, the similarity determination unit configured to determine the similarity between the target text and at least one historical event information, and the history event information having the highest similarity to the target text are output. It also includes a history event information output unit.

In some embodiments, the historical event information includes participant information and trigger word information. The history event information determination unit further has a condition in which the subject or object of the target triplet matches the participant information of the history event information in the history event information set, and the predicate of the target triplet is the history event. It is determined whether or not the condition matching the trigger word information of the historical event information in the information set is satisfied, and the historical event information satisfies at least one of the above-mentioned conditions related to the target text. Is configured to determine.

In some embodiments, the historical event information includes keywords. The similarity determination unit further acquires a first word set by dividing the target text, and is included in the history event information with respect to the history event information in the at least one history event information. By concatenating each keyword and dividing the concatenated text, a second word set is acquired, and based on the first word set and the second word set, the target text and the historical event information It is configured to determine the degree of similarity.

In a third aspect, an embodiment of the invention comprises equipment comprising one or more processors and a storage device in which one or more programs are stored, said one or more. Provided is a facility for realizing the method described in the embodiment of any one of the first aspects in the one or more processors when the program is executed by the one or more processors.

In a fourth aspect, the embodiment of the present invention is a computer-readable storage medium in which a computer program is stored, and the embodiment of any one of the first aspects when the program is executed by a processor. Provided is a computer-readable storage medium that realizes the method described in 1.

The method and apparatus for generating the information according to the embodiment of the present invention can generate a dependency tree of the target text by performing a dependency parsing on the target text after the target text is received. Then, at least one triplet is acquired by matching at least one preset syntax structure tree with the dependency tree. Finally, one target triplet is determined from the at least one triplet based on the words contained in each triplet in the at least one triplet and the preset weights of the syntactic structure tree matched to obtain the triplet. To do. Since the method and device of this embodiment can select the triplet most related to the event contained in the target text, the extraction accuracy rate of the target triplet is improved.

Other features, objectives and advantages of the present invention will become more apparent by reading the detailed description of the non-limiting examples made with reference to the drawings below.
It is a figure which shows the exemplary system architecture to which one Example of this invention is applicable. It is a flowchart which shows one Example of the method for generating the information which concerns on this invention. It is a structural schematic diagram which shows the dependency tree of one Example of the method for generating the information which concerns on this invention. It is a structural schematic diagram which shows the syntactic structure tree of one Example of the method for generating the information which concerns on this invention. FIG. 5 is a structural schematic diagram showing one candidate triplet obtained by matching the dependency tree shown in FIG. 2a with the syntactic structure tree shown in FIG. 2b in the method for generating information according to the present invention. FIG. 5 is a structural schematic diagram showing another candidate triplet obtained by matching the dependency tree shown in FIG. 2a with the syntactic structure tree shown in FIG. 2b in the method for generating information according to the present invention. FIG. 5 is a structural schematic diagram showing another candidate triplet obtained by matching the dependency tree shown in FIG. 2a with the syntactic structure tree shown in FIG. 2b in the method for generating information according to the present invention. It is a schematic diagram which shows one application scenario of the method for generating the information which concerns on this invention. It is a flowchart which determines a target triplet in the method for generating the information which concerns on this invention. It is a flowchart which shows another Example of the method for generating the information which concerns on this invention. It is a structural schematic diagram of one Example of the apparatus for generating the information which concerns on this invention. It is a structural schematic diagram of the computer system applied to the equipment for achieving the Example of this invention.

Hereinafter, the present invention will be described in more detail with reference to the drawings and examples. The steps to be understood are that the specific examples described herein are merely for explaining the related invention and are not intended to limit the invention. For convenience of explanation, only the parts related to the invention are shown in the drawings.

It should be noted that the examples of the present invention and the features in the examples can be combined with each other as long as there is no contradiction. Hereinafter, the present invention will be described in detail with reference to the drawings and examples.

FIG. 1 shows an exemplary system architecture 100 to which examples of methods for generating information or devices for generating information according to the present invention can be applied.

As shown in FIG. 1, system architecture 100 may include terminal devices 101, 102, 103, network 104 and server 105. The network 104 is used to provide a medium for communication links between the terminal devices 101, 102, 103 and the server 105. The network 104 may include various types of connections such as wired, wireless communication links or fiber optic cables.

The user can use the terminal devices 101, 102, 103 to exchange information with the server 105 via the network 104 in order to send and receive messages. Various communication client applications such as text input applications, web browser applications, shopping applications, search applications, instant communication tools, mailbox clients, and social platform software can be installed on the terminal devices 101, 102, and 103.

The terminal devices 101, 102, 103 may be hardware or software. When the terminal devices 101, 102, and 103 are hardware, they may be various electronic devices having a display screen and supporting text input, such as a smartphone, a tablet computer, an electronic book reader, and MP3 (Moving Picture Experts). Group Audio Layer III, Moving Picture Experts Group Audio Layer 3) Player, MP4 (Moving Computer Experts Group Audio Layer IV, Moving Picture Experts Group Audio Layer 4) Player, Laptop Portable Includes, but is not limited to, computers, desktop computers, and the like. When the terminal devices 101, 102, and 103 are software, they may be installed in the above-mentioned electronic device. It may be implemented as multiple software or software modules (eg, used to provide distributed services) or as a single software or software module. Here, it is not particularly limited.

The server 105 may be a server that provides various services, and may be, for example, a back-end server that supports the text entered on the terminal devices 101, 102, 103. The back-end server can perform processing such as analysis on the received data such as target text, and further feed back the processing result (for example, target triplet) to the terminal devices 101, 102, 103.

The server 105 may be hardware or software. When the server 105 is hardware, it may be realized as a distributed server cluster composed of a plurality of servers, or may be realized as a single server. When the server 105 is software, it may be implemented as multiple software or software modules (eg, used to provide distributed services) or as a single software or software module. There is no particular limitation here.

The method for generating information according to the embodiment of the present invention may be executed by the terminal devices 101, 102, 103, or may be executed by the server 105. Correspondingly, the device for generating information may be installed in the terminal devices 101, 102, 103, or may be installed in the server 105.

It should be understood that the system architecture 100 may not include the network 104 and the server 105 if the method for generating information according to the embodiments of the present invention is performed by the terminal devices 101, 102, 103.

It should be understood that the number of terminal devices, networks and servers in FIG. 1 is only exemplary. If necessary, the number of terminal devices, networks, and servers may be arbitrarily adjusted.

Next, reference is made to FIG. 2 showing a flow 200 of an embodiment of the method for generating information according to the present invention. The method for generating the information of the embodiment includes step 201, step 202, step 203 and step 204.

In step 201, the target text is received.

In an embodiment of the invention, the executing entity of the method for generating information (eg, terminal devices 101, 102, 103 or server 105 shown in FIG. 1) can receive the target text. When the executing body of the method for generating information is a terminal device, the target text input by the user via the terminal device can be directly received. When the execution subject of the method for generating information is a server, the target text can be received from the terminal device used by the user by the wired connection method or the wireless connection method. The target text may include an object and descriptive information about the object. The object may be any entity recognized by NER technology or EL technology, and examples thereof include a person, a company, and the like. The descriptive information may be information for explaining the object, and includes, but is not limited to, information for explaining the state of the object, information for explaining the operation of the object, and the like.

The wireless connection method includes 3G / 4G connection, WiFi connection, bluetooth connection, WiMAX connection, Zigbee connection, UWB (ultra wideband) connection, and other currently known or future wireless connection methods. It may include, but is not limited to.

In step 202, the dependency tree of the target text is generated by performing the dependency parsing on the target text.

After receiving the target text, the executor can perform dependent parsing on the target text. Dependency syntax is also called dependency syntax, and is a French linguist, L.A. First submitted by Tenniere in the 1950s of the 20th century. Dependency syntax is a structural syntax that describes the linguistic structure of a sentence using the dependencies formed between words. In order to explain the structural features of the dependency syntax in an easy-to-understand manner, it can be expressed using a dependency tree. Each node in the dependency tree corresponds to a word in the sentence. The dependency tree can not only represent the dependency between words, but can also represent the part of speech of the word (eg, quantifier, particle, etc.), and the function of the word in the text (eg, adnominal modifier, adverbial modifier, etc.). ) Can also be further expressed. In practical applications, the executable can perform dependent parsing on the target text using various open source toolkits. The open source tool kit may include Standford Parser, which is an open source tool kit provided by the Standford NLP group of Stanford University in the United States, and Fudan NLP, which is an open source tool kit developed by the Computer Science Academy of Fudan University in China.

In step 203, at least one triplet is acquired by matching at least one preset syntax structure tree with the dependency tree.

Once the target text dependency tree is generated, the executor can match the preset syntax structure tree with the dependency tree. The tree-like structure of the syntactic structure tree may include a plurality of nodes, and the syntactic structure tree may include the part of speech of the word located at each node. By matching the syntax structure tree with the dependency tree, words having the same dependency as the syntax structure tree in the dependency tree can be obtained. At the same time, the part of speech of each acquired word is the same as the part of speech of the word of the corresponding node in the syntactic structure tree.

As an example, FIG. 2a shows the structure of the target text dependency tree, and FIG. 2b shows the structure of the syntax structure tree. The syntactic structure tree shown in FIG. 2b shows the part of speech of the word of each node, however, v. Represents a verb, n. Represents a noun. In some selective embodiments, the syntactic structure tree and the dependency tree can be matched as follows: First, a candidate triplet consisting of words located at the dashed node in FIGS. 2c, 2d, and 2e can be determined by considering only the structure of the syntactic structure tree and the dependency tree without considering the part of speech of the word. .. Next, the part of speech of the word of each node in the candidate triplet shown in FIGS. 2c, 2d and 2e is matched with the part of speech of the word of each node in the syntactic structure tree, and the part of speech and syntax of each word of the triplet shown in FIG. 2c are matched. It can be determined that the part of speech of each node's word in the structure tree is the same. Therefore, the triplet shown in FIG. 2c is the result of matching the dependency tree and the syntactic structure tree.

The triplet may include a subject, a predicate and an object, and the triplet here may be a triplet in a broad sense. For example, if some words do not have an object, the object in the resulting triplet is "null". For example, some words may contain parallel predicates, and the predicates in the resulting triplet may contain two words. The subject, predicate and object in the triplet may be the same as or different from the subject, predicate and object of the target text. As an example, if the target text is "The share cycle industry is expanding rapidly in the first half of 2016 and 2017 and then gradually declining in the second half of 2017", the triplets obtained. May include "share cycle industry / expanding / null, share cycle industry / decreasing trend / showing". The subject of the target text is "share cycle industry", the predicate is "indicate", and the object is "declining trend". The predicate "extend" in the first triplet is different from the predicate "indicate" in the target text. The subject, predicate and object in the second triplet are the same as the subject, predicate and object of the target text.

In step 204, one target triplet is determined from at least one triplet based on the words contained in one triplet in at least one triplet and the preset weights of the triplet's syntactic structure tree obtained by matching.

When the at least one triplet is obtained, for each triplet of the at least one triplet, the executing entity is based on the words contained in the triplet and the preset weights of the syntax structure tree of the triplet obtained by matching. Therefore, one target triplet can be determined from at least one triplet. The weight of the syntactic structure tree can be set by an engineer according to a specific application scenario. For example, the technician selects one syntax structure tree from at least one syntax structure tree each time to perform triplet matching, and the number of times the syntax structure tree is selected for triplet matching in the past time range. Weights can be set based on. Alternatively, the technician can set the weights based on the number of nodes contained in the syntactic structure tree.

Next, refer to FIG. 3, which is a schematic diagram of an application scenario of the method for generating information according to this embodiment. In the application scenario of FIG. 3, the target text input by the user from the terminal is a video title, and the terminal transmits the video title to the server. Upon receiving the video title, the server can first generate a dependency tree for the video title. It then gets at least one triplet by matching it with the syntactic structure tree. Next, one target triplet is determined from at least one triplet. Finally, the target triplet is output to the terminal so that the user can see it.

In the method for generating information according to the embodiment of the present invention, when the target text is received, the dependency parsing is performed on the target text to generate the dependency tree of the target text. Then, at least one triplet is acquired by matching at least one preset syntax structure tree with the dependency tree. Finally, one target triplet is determined from the at least one triplet based on the words contained in each triplet in the at least one triplet and the preset weights of the syntactic structure tree matched to obtain the triplet. To do. In the method of this embodiment, the triplet most related to the event contained in the target text can be selected, so that the extraction accuracy rate of the target triplet is improved.

In some desired embodiments of this embodiment, the performer can determine the weighting of the syntactic structure tree according to the following steps not shown in FIG. First, get at least one historical target triplet. Next, the number of historical target triplets obtained by matching a predetermined syntactic structure tree among the at least one historical target triplet is statistic. Finally, the weight of at least one syntactic tree is determined based on the statistical results.

In the present embodiment, the execution subject can first acquire at least one history target triplet. Here, the historical target triplet refers to a target triplet obtained by processing a target text received by an executing entity within a past time range. Then, the executing subject can statistic the number of historical target triplets obtained by matching a predetermined syntactic structure tree among the at least one historical target triplet. It is understood that the greater the number of historical target triplets obtained by matching a syntactic tree, the higher the accuracy of the syntactic tree, and thus the greater the weight of the syntactic tree. Should. Finally, the executor can determine the weight of each syntactic tree based on the results of the statistics. As an example, the executor obtains 100 historical target triplets, and from the statistical results, 50 historical target triplets are obtained from the syntactic structure tree a, and 30 historical target triplets are obtained from the syntactic structure tree b. It can be seen that the remaining 20 historical target triplets obtained from were obtained from the syntactic structure tree c. Based on the result of the statistics, the execution subject has a weight of the syntax structure tree a of 50/100 = 0.5, a weight of the syntax structure tree b of 30/100 = 0.3, and a weight of the syntax structure tree c. It can be determined that the weight is 20/100 = 0.2.

In the method for generating the information of the present embodiment, the weight of the syntax structure tree can be adjusted in a timely manner in combination with the historical target triplet, so that the determination accuracy rate of the target triplet is improved.

Next, refer to FIG. 4, which shows the flow 400 for determining the target triplet in the method for generating information according to the present invention. As shown in FIG. 4, the present invention can determine the target triplet based on step 401, step 402, step 403 and step 404.

In step 401, the quantifier and adnominal modifier in the target text are determined based on the dependency tree.

In this embodiment, the dependency tree represents the part of speech and the function of the word, so that the executing subject determines the quantifier and the adnominal modifier in the target text based on the dependency tree of the generated target text. can do. Adnominal modifiers are used to modify the subject and object and may include nouns, pronouns and adjectives.

In step 402, the object to be modified by the quantifier and the object to be modified by the adnominal modifier are determined.

Once the quantifier and the qualifier have been determined, the executing subject can determine the object to be modified by the quantifier and the object to be modified by the quantifier. The object may be the subject in the triplet or the object in the triplet. As an example, in the text "one apple", "one" is a quantifier, and "apple" is an object modified by the quantifier "one". In the text "red apple", "red" is an adnominal modifier, and "apple" is an object modified by the adnominal modifier "red".

In step 403, at least one triplet is updated based on the determined quantifier, adnominal modifier, and object.

The executing subject can update at least one triplet after the quantifier, the adnominal modifier, and the object to be modified are determined. For example, if the determined object is the object of the triplet, the executor combines the quantifier and / or modifier that modifies the object with the object, and the combined text is the new object of the triplet. By doing so, the triplet is updated. With the update, the number of words in each triplet can be increased, and the target triplet can be determined by the number of characters included in the updated triplet, so that the determination accuracy rate of the target triplet can be improved. As an example, the target text is "Mr. Zhangsan attends a birthday party in Shenzhen", and the executing entity creates a dependency tree and then matches it with the syntax structure tree to match the triplet "Mr. Zhangsan"・ You can get "Shenzhen / Attend" and triplet "Mr. Zhangsan / Birthday party / Attend". Based on the dependency tree, it can be determined that "Shenzhen" is an adnominal modifier of "birthday party", and the executing body gets the triplet "Zhangsan-san / Shenzhen's birthday party / attendance" by updating. can do.

In some selective embodiments of this example, step 403 may further include the following content not shown in FIG. The triplet in which the object and the adnominal modifier of the target text are matched in the at least one triplet is deleted.

In the present embodiment, the executing subject can determine whether or not a triplet whose object is an adnominal modifier of the target text exists in at least one triplet obtained. If present, the executing entity can delete the triplet. For example, for the triplet "Mr. Zhang San, Shenzhen, attend", the executing subject can determine that the adnominal modifier "Shenzhen" should not be the object of the triplet. Therefore, the executing entity can determine that the triplet "Mr. Zhang San, Shenzhen, Attend" is incorrect and delete the triplet. Therefore, the amount of calculation can be effectively reduced, and the efficiency of calculation can be improved.

In some selective embodiments of this embodiment, the performer can update the triplet according to the following steps not shown in FIG. First, for the triplet of at least one triplet, it is determined whether or not the determined object matches the subject or object of the triplet. Next, after it is determined that the determined object matches the subject of the triplet, the quantifier, the modifier and the subject of the triplet that modify the determined object are combined, and the combined text is the triplet's subject. Determined as the subject. Then, after it is determined that the determined object matches the object of the triplet, the quantifier, the modifier and the object of the triplet that modify the determined object are combined, and the combined text is combined with the triplet. Determined as the object of.

For each triplet of the at least one triplet, the executing subject can first determine whether or not the determined object matches the subject or object of the triplet. It should be understood here that matching may mean that at least one character in the object is the same as at least one character in the subject or object of the triplet. For example, when the object is "Zhang-san" and the subject of the triplet is "Zhang-san", it can be determined that the object matches the subject of the triplet.

After it is determined that the determined object matches the subject of the triplet, the executing subject combines the quantifier, the modifier and the subject of the triplet that modify the object, and the combined text is combined with the subject of the triplet. can do. For example, if the object is "Zhang-san", the adnominal modifier that modifies the object is "refreshing", and the subject of the triplet is "Zhang-san", the combined text will be "refreshing Zhang-san". May be. After that, "refreshing Mr. Zhangsan" is used as the subject of the triplet. This provided an update to the triplet subject.

After it is determined that the determined object matches the object of the triplet, the executing subject combines the quantifier, the modifier and the object of the triplet that modify the object, and the combined text is the object of the triplet. Can be an object. This resulted in an update to the triplet object.

It is understood that when updating a triplet, only the triplet subject may be updated, only the triplet object may be updated, or the triplet subject and object may be updated at the same time. I want to. Further, when performing the combination operation, any one of the quantifier and the adnominal modifier may be combined with the subject of the triplet or the object of the triplet.

In step 404, one target triplet is determined from at least one updated triplet.

After the triplet has been updated, one target triplet can be determined from at least one updated triplet. Specifically, the executing entity can determine the target triplet according to sub-step 4041 and sub-step 4042.

In substep 4041, for the triplet of at least one triplet, the preset weight of the syntactic structure tree of the triplet obtained by matching is determined, the number of characters of the word contained in the triplet is determined, and the above-mentioned The degree of co-occurrence of a word in a triplet is determined, and the score of the triplet is determined based on the determined weight, the number of characters, and the degree of co-occurrence.

For each triplet of at least one triplet, the executing entity can first determine the weight of the syntactic tree of the triplet obtained by matching. Next, the number of characters of the word contained in the triplet is determined based on the word contained in the triplet. Next, the degree of co-occurrence of words in the triplet is determined. Finally, the triplet score is calculated based on the determined weight, number of characters and degree of co-occurrence. The co-occurrence here may mean that the words in the triplet appear in the same phrase, the same paragraph, or the same sentence. The degree of co-occurrence is the probability that the first word in the triplet will appear, the probability that the second word will appear after the first word appears, and the probability that the first word and the second word will appear. It may be the product of the probability that the third word appears in.

For example, when the triplet is "Mr. Zhangsan, newborn baby, visit", the executing entity can first determine the probability that "Mr. Zhangsan" will appear in the preset set of information. The set of information may be a set of homepage titles, a set of a large number of sentences, and the like. Assuming that the set of information contains 10,000 messages, and the number of messages including "Zhangsan" is 100, the probability that "Zhangsan" appears is 1%. Then, the executing subject can determine the probability that "visit" appears from the message including "Mr. Zhangsan" in the set of information. Assuming that 20 of the 100 messages including "Zhangsan" include "visit", the probability that "visit" will appear after "Zhangsan" appears is 20. %. After that, the executing subject can determine that the probability that "newborn" appears as the object of "visit" is 50% after "Zhangsan" and "visit" appear in the same way. Therefore, the degree of co-occurrence is 1% × 20% × 50% = 0.1%.

After acquiring the weight, the number of characters, and the degree of co-occurrence, the executing subject can determine the score of the triplet according to the following equation. Score = a x weight + b x number of characters + c x co-occurrence degree. In the formula, a, b, and c are preset coefficients.

In substep 4042, the triplet with the highest score among at least one triplet is determined as the target triplet.

After the score of each triplet is obtained, the executing entity can use the triplet having the highest score among the at least one triplet as the target triplet. It should be understood that the higher the triplet score, the higher the accuracy of the triplet and the better the representation of the objects contained in the target text and the descriptive information of the objects.

In the method for generating information according to the above embodiment of the present invention, one triplet most related to the target text can be determined from a plurality of triplets, so that the extraction accuracy rate of the triplet is improved.

Next, refer to FIG. 5, which shows the flow 500 of another embodiment of the method for generating the information of the present invention. As shown in FIG. 5, the method for generating the information of this embodiment can further include steps 501, 502 and 503 after the target triplet is obtained.

In step 501, at least one historical event information associated with the target text in the preset historical event information set is determined based on the target triplet.

After the target triplet is determined, at least one historical event information associated with the target text in the preset historical event information set based on the target triplet can be determined. The historical event information may include an object and descriptive information of the object. In this embodiment, if the object in the historical event information is the same as the subject of the target triplet, or if the historical event information includes the subject, predicate, or object of the target triplet, the historical event information is the target. It can be considered to be related to the text.

In some selective embodiments of this embodiment, the historical event information may include participant information and trigger word information. The executing subject can determine whether or not the historical event information and the target text are related according to the next step (not shown in FIG. 5). First, the condition that the subject or object of the target triplet matches the participant information of the history event information in the history event information set, and the predicate of the target triplet and the trigger word information of the history event information in the history event information set match. Judge whether or not the conditions are met. It is then determined that the historical event information satisfies at least one of the above-mentioned conditions associated with the target text.

In the present embodiment, the participant information may be information on a related person in a historical event. The trigger word information may be the operation information of the participant information. For example, if the history event information is "Mr. Komei and Mr. Koboku go to the first dining room together for lunch", the participant information is "Mr. Komei" and "Mr. Koboku". May include, and the trigger word information is "eat". If the subject or object of the target triplet is matched with the participant information, and if they match, the subject or object of the triplet is considered to be the same as the participant in the history event. If the target triplet predicate and the trigger word information are matched and they match, the triplet predicate is considered to be the same as the trigger word in the historical event. If at least one of the two conditions is met, the executing entity can consider the historical event to be related to the target text.

In step 502, the similarity between the target text and at least one historical event information is determined.

After at least one historical event information has been determined based on the target triplet, in order to further acquire the historical event information most related to the target text, the executing entity is among the target text and the at least one historical event information. The degree of similarity with each historical event information can be determined. The executing entity can determine the similarity between the target text and the historical event information based on the number of the same characters or words in the target text and the historical event information. Alternatively, the executing entity can further determine the similarity with respect to the number of items satisfying the above conditions in the historical event information.

In some selective embodiments of this embodiment, the historical event information may include keywords. The keyword may be an event name, an event occurrence time, or the like. The event name may include the subject, predicate, and object of the history event. The executor can determine the similarity between the target text and the historical event information according to the following steps not shown in FIG. First, the first word set is acquired by dividing the target text. Next, for each history event information in the at least one history event information, each keyword included in the history event information is concatenated, and the concatenated text is divided to form a second word set. get. The similarity between the target text and the historical event information is determined based on the first word set and the second word set.

In the present embodiment, the execution subject can first acquire the first word set by dividing the target text. At the time of division, it may be divided based on the meaning or the number of characters. Next, for each historical event information in the at least one historical event information, the executing subject concatenates each keyword included in the historical event information and divides the concatenated text to obtain a second. Get the word set of. It can be divided at the same particle size to ensure the correctness of similarity. That is, when dividing the target text and the combined text, both are divided using the method of bigram or trigram, and the number of characters contained in the obtained word is the same. For example, if the target text is "I am Chinese", splitting it using the bigram method will give "I", "Hachu", "China", and "Kokujin", but using the trigram method. If you divide it, you will get "I am medium", "is Chinese", and "Chinese".

The executing entity can enumerate all the words in the first word set and the second word set after the first word set and the second word set are obtained. After that, the number of times each word appears in the target text is statistic, and the obtained times are combined to obtain the first word vector A. Next, the number of times each word appears in the combined text is statistic, and the obtained times are combined to obtain a second word vector B. The executor then calculates the similarity of the text combined with the target text based on the vector cosine formula.

In the equation, A = (A ₁ , A ₂ , ..., _An ), B = (B ₁ , B ₂ , ..., B _n ). However, A _i is the i th value of the first word vectors A, B _i is the i th value of the second word vector B.

In step 503, the historical event information having the highest degree of similarity to the target text is output.

After the similarity between each historical event information associated with the target text and the target text is determined, the execution subject can output the historical event information having the highest similarity with the target text.

In the method for generating the information according to the embodiment of the present invention, the historical event information most related to the target text in the historical event information set can be determined in order to enhance the amount of information of the user. The method of this embodiment can be applied to screen videos. By using the title of the video as the target text, determining the target triplet of the video title, and then selecting the historical events associated with the video title, it is possible to determine whether the video is an old video.

Further, as shown in FIG. 6, as a realization of the method shown in each of the drawings, the present invention provides an embodiment of an apparatus for generating information. The embodiment of the apparatus corresponds to the embodiment of the method shown in FIG. 2, and the apparatus can be specifically applied to various electronic devices.

As shown in FIG. 6, the device 600 for generating the information of this embodiment includes a target text receiving unit 601, a dependency tree generating unit 602, a triplet determining unit 603, and a target triplet determining unit 604.

Here, the target text receiving unit 601 is configured to receive the target text. The target text contains the object and descriptive information for the object.

The dependency tree generation unit 602 is configured to perform dependency parsing on the target text and generate a dependency tree for the target text.

The triplet determination unit 603 is configured to acquire at least one triplet composed of a subject, a predicate, and an object by matching at least one preset syntactic structure tree with a dependency tree.

The target triplet determination unit 604 has one target triplet from at least one triplet based on the preset weights of the syntactic structure tree that are matched to obtain the words and triplets contained in one triplet in at least one triplet. Is configured to determine.

In some selective embodiments of this embodiment, the target triplet determination unit 604 further comprises a coalition modifier determination module, an object determination module, a triplet update module and a target triplet determination module not shown in FIG. It may be.

The adnominal modifier determination module is configured to determine the quantifiers and adnominal modifiers in the target text based on the dependency tree.

The object determination module is configured to determine the object to be modified by the quantifier and the object to be modified by the adnominal modifier.

The triplet update module is configured to update at least one triplet based on the determined quantifiers, adnominal modifiers and objects.

The target triplet determination module is configured to determine one target triplet from at least one updated triplet.

In some selective embodiments of this embodiment, the triplet update module determines whether the determined object matches the subject or object of the triplet for the triplet of at least one triplet. In response to the determination that the determined object matches the subject of the triplet, the quantifier, the modifier and the subject of the triplet that modify the determined object are combined, and the combined text is obtained. In response to the determination as the subject of the triplet and the determination that the determined object matches the object of the triplet, the quantifier, the coalition modifier and the object of the triplet that modify the determined object are combined. , The combined text may be further configured to determine the subject of the triplet.

In some selective embodiments of this embodiment, the target triplet determination module presets the syntactic structure tree that is matched against the triplet of at least one triplet to obtain the triplet. The weight is determined, the number of characters of the word contained in the triplet is determined, the degree of co-occurrence of the word in the triplet is determined, and the score of the triplet is determined based on the determined weight, the number of characters and the degree of co-occurrence. The triplet with the highest score among at least one triplet may be further configured to be determined as the target triplet.

In some selective embodiments of this embodiment, the device 600 may further include a weighting unit not shown in FIG. The weighting unit may include a history target triplet module, a triplet quantity statistics module, and a weight determination module.

The history target triplet module is configured to acquire at least one history target triplet.

The triplet quantity statistics module is configured to stat the number of historical target triplets obtained by matching from a predetermined syntactic structure tree among the at least one historical target triplet.

The weight determination module is configured to determine the weight of at least one syntactic structure tree based on the result of statistics.

In some selective embodiments of this embodiment, the device 600 may further include a history event information determination unit, a similarity determination unit and a history event information output unit not shown in FIG.

Here, the historical event information determination unit is configured to determine at least one historical event information associated with the target text in the preset historical event information set based on the target triplet.

The similarity determination unit is configured to determine the similarity between the target text and at least one historical event information.

The history event information output unit is configured to output history event information having the highest degree of similarity to the target text.

In some selective embodiments of this embodiment, the historical event information may include participant information and trigger word information. The history event information determination unit further includes a condition in which the subject or object of the target triplet matches the participant information of the history event information in the history event information set, and the predicate of the target triplet is the history event information in the history event information set. It is configured to determine whether or not a condition that matches the trigger word information is satisfied, and to determine that the historical event information satisfies at least one of the above-mentioned conditions associated with the target text.

In some selective embodiments of this embodiment, the historical event information may include keywords. The similarity determination unit further acquires a first word set by dividing the target text, and for the historical event information of at least one historical event information, each keyword included in the historical event information. Is concatenated and the concatenated text is divided to obtain a second word set, and the similarity between the target text and the history event information is determined based on the first word set and the second word set. It is configured as follows.

The device for generating the information according to the embodiment of the present invention can generate a dependency tree of the target text by performing a dependency parsing on the target text after the target text is received. Next, at least one triplet is acquired by matching at least one preset syntax structure tree with the dependency tree. Finally, one target triplet is determined from the at least one triplet based on the words contained in each triplet in the at least one triplet and the preset weights of the syntactic structure tree matched to obtain the triplet. To do. Since the device of this embodiment can select the triplet most related to the event contained in the target text, the extraction accuracy rate of the target triplet is improved.

It should be understood that the units 601 to 604 described in the device 600 for generating information correspond to each step in the method shown in FIG. Therefore, the operations and features described above for the method for generating information are also suitable for the device 600 and the units contained therein, and will not be described further here.

Hereinafter, FIG. 7 which is a structural schematic diagram showing a computer system 700 applied to the equipment for realizing the embodiment of the present invention will be referred to. The equipment shown in FIG. 7 is merely an example, and does not limit the functions and the range of use of the embodiments of the present invention.

As shown in FIG. 7, the computer system 700 has various appropriate operations and processes depending on the program stored in the read-only memory (ROM) 702 or the program loaded from the storage unit 708 into the random access memory (RAM) 703. 701 is provided with a central processing unit (CPU) 701 capable of executing the above. The RAM 703 further stores various programs and data necessary for the operation of the system 700. The CPU 701, ROM 702, and RAM 703 are connected to each other via the bus 704. The input / output (I / O) interface 705 is also connected to the bus 704.

Input unit 706 including keyboard, mouse, etc., cathode ray tube (CRT), liquid crystal display (LCD), etc., output unit 707 including speakers, storage unit 708 including hard disk, etc., and network interface including LAN card, modem, etc. The communication unit 709 of the card is connected to the I / O interface 705. The communication unit 709 executes communication processing via a network such as the Internet. The driver 710 is connected to the I / O interface 705 as needed. The removable media 711 is, for example, a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, or the like, and is attached to a driver 710 as needed, thereby reading a computer program from the removable media 711. Is installed in the storage unit 708 as needed.

In particular, according to the embodiment of the present invention, the process described with reference to the above flowchart may be realized as a software program of a computer. For example, an embodiment of the present invention comprises a computer program product comprising a computer program embodied in a machine-readable storage medium, said computer program comprising program code for performing the method shown in the flow chart. In such an embodiment, the computer program may be downloaded and installed from the network via the communication unit 709 and / or installed from the removable media 711. When the computer program is executed by a central processing unit (CPU) 701, it performs the above-mentioned functions limited by the method of the present invention.

The computer-readable storage medium of the present invention may be a computer-readable signal storage medium, a computer-readable storage medium, or any combination thereof. The computer-readable storage medium can be, but is not limited to, for example, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices or devices, or any combination thereof. More specific examples of computer-readable storage media include electrical connections with one or more leads, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), and erasable programmable read-only memory. It can include, but is not limited to, (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical memory, magnetic memory, or any suitable combination thereof.

In the present invention, the computer-readable storage medium may be any tangible storage medium that includes or stores programs that can be used by, or incorporated into, instruction execution systems, devices or devices. In the present invention, the computer-readable signal storage medium may include a data signal carrying a computer-readable program code that is propagated within the baseband or as part of a carrier. Such propagated data signals can take various forms and can include, but are not limited to, electromagnetic signals, optical signals, or any suitable combination thereof. The computer-readable signal storage medium may be any computer-readable storage medium other than the computer-readable storage medium. The computer-readable storage medium can transmit, propagate or transfer programs used by or embedded in instruction execution systems, devices or devices. The program code contained in the computer-readable storage medium can be transmitted on any suitable medium, including, but not limited to, wireless, wired, optical cable, RF, etc., or any suitable combination thereof.

The computer program code for performing the operation of the present invention can be created in one or more programming languages or a combination thereof, and the programming language is an object such as Java (registered trademark), Smalltalk, C ++, etc. Includes oriented programming languages and traditional procedural programming languages such as the "C" language or similar programming languages. The program code runs entirely on the user's computer, partially on the user's computer, as a separate software package, partly on the user's computer, and partly on the remote computer. , Or may be run entirely on a remote computer or server. When involved with a remote computer, the remote computer can be connected to the user's computer via any type of network, including a local area network (LAN) or wide area network (WAN), or is connected to an external computer. Can be (eg, connected to the internet by an internet service provider).

The flowcharts and block diagrams in the drawings show the architecture, functions, and operations that can be realized by the system, method, and computer program product according to each embodiment of the present invention. Here, each block in the flowchart or block diagram may represent one module, program segment or part of code, and the module, program segment or part of code is for achieving a defined logic function. Contains one or more executable instructions of. It should be noted that in some alternative embodiments, the functions shown within the block may be performed in a different order than shown in the drawings. For example, two blocks shown in succession may actually be executed approximately in parallel or in reverse order, depending on the associated function. In addition, each block in the block diagram and / or the flowchart, and the combination of the blocks in the block diagram and / or the flowchart may be realized by a dedicated hardware-based system that executes a specified function or operation, or may be realized. , May be executed in combination with dedicated hardware and computer instructions.

The units described in the examples of the present invention may be realized by software or hardware. The described units may be provided in the processor, for example, "the processor includes a target text receiving unit, a dependency tree generation unit, a triplet determination unit, and a target triplet determination unit". Here, the names of these units do not limit the unit itself in some cases, and for example, the target text receiving unit may be described as "a unit that receives the target text".

On the other hand, the present invention further provides a computer-readable storage medium, and the computer-readable storage medium may be included in the apparatus described in the above embodiment, and may exist independently and be assembled in the apparatus. It may not be. The computer-readable storage medium carries one or more programs, and when the one or more programs are executed by the device, the device includes an object and a target containing descriptive information about the object. It receives the text, performs a dependency parsing on the target text to generate a dependency tree for the target text, matches at least one preset syntax structure tree with the dependency tree, and matches the subject, predicate, and dependency tree. At least one triplet based on the preset weights of the parse structure tree that gets at least one triplet of objects and is matched to get the words and triplets contained in one triplet in at least one triplet. Determine one target triplet from.

The above description is merely a description of the preferred embodiments of the present invention and the technical principles used. The scope of the invention according to the present invention is not limited to the technical proposal consisting of a specific combination of the above technical features, and the above technical features or their technical features are not deviated from the above-mentioned gist of the present invention. Those skilled in the art should understand that other technical proposals consisting of any combination of equivalent features should also be included. For example, there is a technical proposal that replaces the above-mentioned features with the technical features having similar functions disclosed in the present invention (not limited to these).

Claims (19)

A computer-implemented method of generating information,
A step of executing a process of receiving an object and a target text including descriptive information for the object, and
A step of executing a process of performing dependency parsing on the target text and generating a dependency tree of the target text, in which each node of the dependency tree corresponds to each word of the target text. When,
A step of matching at least one preset syntactic structure tree with the dependency tree to acquire at least one triplet consisting of a subject, a predicate, and an object, which is a step of executing the process of obtaining at least one triplet of the syntactic structure tree. Each node contains the part of speech of the word located at that node, and the triplet consists of a step consisting of nodes in the dependency tree that match the syntax structure tree .
For each of said at least one triplet of words contained in the triplet characteristic, and, based on a pre-set weight of the syntactic structure tree that is matched to obtain the triplet, the predetermined rule, of the triplets It is a step of determining a score and executing a process of determining one target triplet from the at least one triplet based on the determined score, and the score is an object included in the target text and a description of the object. Steps that evaluate the accuracy of expressing information ,
A step of executing a process of determining at least one historical event information related to the target text in a preset historical event information set based on the target triplet.
Including methods. The process of determining one target triplet from at least one triplet is
The step of determining the quantifier and the adnominal modifier in the target text based on the dependency tree,
A step of determining an object to be modified by the quantifier and an object to be modified by the adnominal modifier.
A step of updating the at least one triplet based on the determined quantifier, adnominal modifier and object.
The step of determining the one target triplet from the at least one updated triplet, and
The method according to claim 1. The step of updating the at least one triplet based on the determined quantifier, adnominal modifier and object is
A step of determining whether or not the determined object matches the subject or the object of the triplet with respect to the triplet of the at least one triplet.
In response to the determination that the determined object matches the subject of the triplet, the quantifier, the coalition modifier and the subject of the triplet that modify the determined object are combined. With the step of determining the combined text as the subject of the triplet,
In response to the determination that the determined object matches the object of the triplet, the quantifier, the coalition modifier, and the object of the triplet that modify the determined object are added. With the step of combining and determining the combined text as the object of the triplet,
2. The method according to claim 2. The process of determining one target triplet from at least one triplet is
For the triplet of at least one triplet, the preset weight of the syntactic structure tree matched to obtain the triplet is determined, and the number of characters of the word contained in the triplet is determined. , A step of determining the degree of co-occurrence of the word contained in the triplet, and determining the score of the triplet based on the determined weight, the number of characters, and the degree of co-occurrence.
The step of determining the triplet with the highest score among the at least one triplet as the target triplet, and
The method according to any one of claims 1 to 3. The method is
A step to execute the process of acquiring at least one historical target triplet, and
A step of executing a process of statisticizing the number of historical target triplets obtained by matching a predetermined syntactic structure tree among the at least one historical target triplets.
A step of executing a process of determining the weight of at least one syntactic structure tree based on the statistical result,
The method according to any one of claims 1 to 3, further comprising. The method,
And executing a process of determining the similarity of the said target text at least one historical event information,
A step of executing a process of outputting historical event information having the highest degree of similarity to the target text, and
The method according to claim 1, further comprising. The historical event information includes participant information and trigger word information.
The process of executing the process of determining at least one historical event information related to the target text in the preset historical event information set based on the target triplet is
The condition that the subject or object of the target triplet matches the participant information of the history event information in the history event information set, or the predicate of the target triplet is the history event information in the history event information set. A step of determining whether or not the condition matching the trigger word information is satisfied, and
A step of determining that the historical event information satisfies at least one of the above-mentioned conditions associated with the target text.
The method according to claim 1. The historical event information includes keywords and includes keywords.
The process of determining the similarity between the target text and the at least one historical event information is
The step of dividing the target text to obtain the first word set, and
A second word set is acquired by concatenating each keyword included in the history event information with respect to the history event information of the at least one history event information and dividing the concatenated text. Steps and
A step of determining the similarity between the target text and the historical event information based on the first word set and the second word set.
6. The method according to claim 6. A target text receiving unit configured to perform a process of receiving an object and a target text containing descriptive information about the object, and a target text receiving unit.
It is a dependency tree generation unit configured to perform a dependency parsing on the target text and generate a dependency tree of the target text, and each node of the dependency tree is the said. Dependency tree generation unit for each word in the target text ,
Matching the at least one syntax structure tree and the dependency tree is set in advance, subject, met triplet determination unit configured to perform a process of acquiring the predicate and at least one triplet consists object Each node of the syntactic structure tree contains the part of speech of the word located at the node, and the triplet is a triplet determination unit composed of nodes of the dependency tree matching the syntactic structure tree .
For each of said at least one triplet of words contained in the triplet characteristic, and, based on a pre-set weight of the syntactic structure tree that is matched to obtain the triplet, the predetermined rule, of the triplets A target triplet determination unit configured to determine a score and perform a process of determining one target triplet from the at least one triplet based on the determined score, wherein the score is the target text. The target triplet determination unit, which evaluates the accuracy of the objects contained in the lens and the descriptive information of the objects, and the target triplet determination unit .
A historical event information determination unit configured to perform a process of determining at least one historical event information associated with the target text in a preset historical event information set based on the target triplet.
A device for generating information including. The target triplet determination unit is
An adnominal modifier determination module configured to determine quantifiers and adnominal modifiers in the target text based on the dependency tree.
An object determination module configured to determine the object to be modified by the quantifier and the object to be modified by the adnominal modifier.
A triplet update module configured to update at least one triplet based on the determined quantifier, adnominal modifier and object.
A target triplet determination module configured to determine the one target triplet from the at least one updated triplet.
9. The apparatus according to claim 9. The triplet update module further determines, for the triplet of at least one triplet, whether the determined object matches the subject or object of the triplet.
In response to the determination that the determined object matches the subject of the triplet, the quantifier, the coalition modifier and the subject of the triplet that modify the determined object are combined. The combined text is determined as the subject of the triplet and
In response to the determination that the determined object matches the object of the triplet, the quantifier, the coalition modifier, and the object of the triplet that modify the determined object are added. Combined and configured to determine the combined text as the object of the triplet.
The device according to claim 10. The target triplet determination unit further determines, for the triplet of at least one triplet, the preset weight of the syntactic structure tree that is matched to obtain the triplet, and includes the triplet. The number of characters of the word to be used is determined, the degree of co-occurrence of the word contained in the triplet is determined, and the score of the triplet is determined based on the determined weight, number of characters and degree of co-occurrence.
Of the at least one triplet, the triplet with the highest score is configured to be determined as the target triplet.
The device according to any one of claims 9 to 11. The device is
A historical target triplet module configured to retrieve at least one historical target triplet,
A triplet quantity statistics module configured to statistic the number of historical target triplets obtained by matching a given syntactic structure tree of at least one historical target triplet.
A weight determination module configured to determine the weight of at least one syntax structure tree based on the statistical results.
Including a weight installation unit consisting of
The device according to any one of claims 9 to 11. The apparatus comprising
A similarity determination unit configured to perform a process of determining the similarity between the target text and at least one historical event information.
It further includes a history event information output unit configured to execute a process of outputting history event information having the highest degree of similarity to the target text.
The device according to claim 9. The historical event information includes participant information and trigger word information.
In the history event information determination unit, a condition in which the subject or object of the target triplet matches the participant information of the history event information in the history event information set, or a predicate of the target triplet is the history. It is determined whether or not the condition that matches the trigger word information of the history event information in the event information set is satisfied.
The historical event information is configured to determine that it meets at least one of the above-mentioned conditions associated with the target text.
The device according to claim 9. The historical event information includes keywords and includes keywords.
The similarity determination unit further divides the target text to obtain a first word set.
A second word set is acquired by concatenating each keyword included in the history event information with respect to the history event information of the at least one history event information and dividing the concatenated text. , The similarity between the target text and the historical event information is determined based on the first word set and the second word set.
The device according to claim 14. It ’s a facility,
With one or more processors
A storage device in which one or more programs are stored,
The method according to any one of claims 1 to 8, when the one or more programs are executed by the one or more processors, causes the one or more processors to execute the method.
Facility. A computer-readable storage medium that stores computer programs.
The method according to any one of claims 1 to 8 is realized when the program is executed by a processor.
Computer-readable storage medium. It ’s a computer program,
The method according to any one of claims 1 to 8 is realized when the computer program is executed by a processor.
Computer program.

Images